[0001] The present disclosure concerns a method and apparatus for monitoring abrasive machining.
[0002] Abrasive machining methods such as grinding, deburring and linishing are used in
the final stages of component manufacture to achieve a desired finish, geometry or
to remove unwanted features or artefacts from a component. Machining may be interrupted
to inspect the component, for example to measure the surface roughness or component
geometry.
[0003] According to a first aspect there is provided a method of monitoring abrasion of
a component during abrasive machining, the method comprising: generating an acceleration
signal corresponding to movement of the component or of an abrasive tool in response
to abrasive machining of the component by the tool; generating an acoustic emission
signal corresponding to acoustic waves emitted from within the component; and determining
an abrasion parameter relating to the removal of material from the component based
on a set of input variables including the acceleration signal and the acoustic emission
signal.
[0004] The acceleration signal may be generated by an accelerometer coupled to the component.
The accelerometer may be a multi-axis accelerometer configured to generate an acceleration
signal corresponding to an absolute magnitude of acceleration independent of acceleration
direction.
[0005] The acceleration signal may be generated by an accelerometer coupled to the abrasive
tool. For example, the accelerometer may be a single-axis accelerometer. For example,
if the abrasive tool is a rotary tool, the accelerometer may be coupled to the abrasive
tool (or a mount therefor) and configured to generate an acceleration signal corresponding
to acceleration along a direction orthogonal to the rotational axis of the tool. The
accelerometer may be configured to generate an acceleration signal corresponding to
a magnitude of acceleration in a plane normal to the rotational axis of the tool.
[0006] The method may further comprise: generating a force signal corresponding to a force
acting between the abrasive tool and the component. The set of input variables may
include the force signal. The force signal may be generated by a force sensor coupled
to the tool.
[0007] The abrasion parameter may be selected from the group consisting of: a material removal
parameter relating to the rate at which material is removed from the component by
abrasion; a chamfer dimension parameter relating to a dimension of a chamfer formed
by abrasion; and a surface roughness parameter relating to a surface roughness of
the component.
[0008] The chamfer dimension parameter may relate to the distance by which two surfaces
are separated by the chamfer. The chamfer dimension parameter may relate to the distance
by which two surfaces are separated by the chamfer at the location where the tool
is engaged with the component. It will be appreciated that the chamfer dimension parameter
may vary along an elongate extent of the chamfer.
[0009] The abrasion parameter may be determined by a fuzzy inference module configured to
determine the abrasion parameter by evaluating a plurality of fuzzy logic rules based
on the set of input variables. The rules may be based on each input variable of the
set of input variables. Each rule may be based on each input variable of the set of
input variables. The fuzzy inference module may be configured to sample each signal
in the set of input variables and determine whether a crisp value derived from the
sampled signal falls within each of a plurality of membership functions for the respective
signal. For example, a crisp value may fall within a membership function if it is
within upper and lower bounds for the membership function. The fuzzy inference system
(FIS) may determine a truth value corresponding to how closely the crisp value matches
the membership function. For example, a crisp value corresponding to the median value
within a range defined for the membership function may result in a truth value of
1.
[0010] The fuzzy inference system may be configured to evaluate a plurality of rules based
on the membership functions for each of the input variables, each rule defining whether
a an outcome variable (i.e. the abrasion parameter) lies within one of a plurality
of membership functions for the outcome variable. A truth value for the outcome variable
(in this example the abrasion parameter) may be determined based on the truth variables
determined for the membership functions of the input variables.
[0011] A crisp value of the outcome variable may be determined based on the determined truth
values and membership functions for each of the plurality of rules. For example, a
crisp value may be determined based on a centroid calculation for each of the truth
values of the membership functions.
[0012] According to a second aspect there is provided a method of abrasive machining, comprising:
causing an abrasive tool to engage a component to conduct an abrasive machining operation;
monitoring abrasion of the component in accordance with the first aspect to determine
an abrasion parameter; and controlling the abrasive machining operation based on the
abrasion parameter.
[0013] The abrasive tool may be a compliant abrasive tool. In other words, the tool may
have one or more resilient elements configured to engage a component to abrade the
component.
[0014] At least one of a tool engagement force; a path of relative movement of the abrasive
tool over the component; and a residence time of the tool at a first location of the
component may be controlled based on the abrasion parameter. Controlling the abrasive
machining operation may comprise comparing a value of the abrasion parameter for a
first location of the component determined by monitoring with a target value for the
first location.
[0015] The abrasive machining operation may be selected from the group consisting of: linishing;
deburring; surface grinding; and edge grinding.
[0016] According to a third aspect there is provided an apparatus comprising: a tool mount
for supporting an abrasive tool to engage a component; an accelerometer for generating
an acceleration signal corresponding to movement of the component or movement of the
abrasive tool in response to abrasive machining of the component by the tool; an acoustic
emission sensor for generating an acoustic emission signal corresponding to acoustic
waves emitted from within the component in response to abrasive machining of the component
by the tool; and a controller configured to determine an abrasion parameter relating
to the removal of material from the component based on a set of inputs including the
acceleration signal and the acoustic emission signal. The abrasion parameter may be
determined based on each of the set of inputs.
[0017] The accelerometer and/or the acoustic emission sensor may be configured to be coupled
to the component. In use, the accelerometer and/or the acoustic emission sensor may
be coupled to the component.
[0018] The apparatus may further comprise: a force sensor coupled to the tool mount and
configured to output a force signal corresponding to a force acting between an abrasive
tool supported by the tool mount and a component with which it is engaged in us. The
controller may be configured to determine the abrasion parameter based on a set of
inputs including the force signal. In other words, the abrasive tool may be coupled
to, attached to or received in the tool mount for engaging a component.
[0019] The controller may be configured to determine the abrasion parameter by evaluating
a plurality of fuzzy logic rules based on the set of input variables.
[0020] The tool mount may be for supporting a rotary abrasive tool. The tool mount may be
configured to rotate the rotary abrasive tool. For example, the mount may include
a chuck for supporting a rotary abrasive tool. The apparatus may further comprise
a compliant abrasive tool supported by the tool mount. The compliant abrasive tool
may comprise a rotary compliant abrasive tool, for example a flap-wheel comprising
a plurality of fins of abrasive material extending from a hub. The compliant abrasive
tool may comprise a belt of abrasive material supported on a circulating path, for
example, for linishing, deburring or grinding.
[0021] According to a fourth aspect there is provided a non-transitory machine-readable
storage medium encoded with instructions executable by a processor, including instructions
to: receive an acceleration signal corresponding to movement of a component or movement
of an abrasive tool in response to abrasive machining of the component by the tool;
receive an acoustic emission signal corresponding to acoustic waves emitted from within
the component during abrasive machining of the component; determine an abrasion parameter
relating to the removal of material from the component based on a set of inputs including
the acceleration signal and the acoustic emission signal.
[0022] The instructions may include instructions to receive a force signal corresponding
to a force acting between the abrasive tool and the component during abrasive machining.
The set of inputs may include the force signal.
[0023] The skilled person will appreciate that except where mutually exclusive, a feature
described in relation to any one of the above aspects may be applied
mutatis mutandis to any other aspect. Furthermore except where mutually exclusive any feature described
herein may be applied to any aspect and/or combined with any other feature described
herein.
[0024] Embodiments will now be described by way of example only, with reference to the Figures,
in which:
Figure 1 schematically shows an example apparatus for abrasive machining;
Figure 2 schematically shows an example test component for calibrating abrasive machining
monitoring;
Figure 3 is a plot showing correlated example input variables and an example abrasion
parameter.
Figure 4 is an illustration of membership functions for an acceleration energy signal;
Figure 5 is an illustration of membership functions for an acoustic emission energy
signal;
Figure 6 is a table showing example rules of a fuzzy inference system;
Figure 7 is a flowchart of an abrasive machining and monitoring method;
Figure 8 is a flowchart of a calibration method for abrasive machining monitoring;
Figure 9 shows plots of predicted versus empirical abrasion parameters; and
Figure 10 shows a non-transitory machine readable storage medium and processor.
[0025] Figure 1 shows an example abrasive machining apparatus 100, an example abrasion monitoring
apparatus 200, and an example component 302 supported on a work platform 300.
[0026] In this example, the abrasive machining apparatus 100 comprises a multi-axis robotic
arm 102 having three degrees of freedom. The robotic arm 102 comprises, at its distal
end, a tool mount 104 for supporting an abrasive tool 106. In this particular example,
the abrasive tool 106 is a compliant abrasive tool, in particular a rotary compliant
abrasive tool such as a flap-wheel having a plurality of fins of abrasive material
extending from a rotatable hub.
[0027] In other examples, different abrasive tools may be used and may be supported by other
supporting apparatus. For example, a linishing tool may comprise a band of abrasive
material supported on a circulating path, and such a linishing tool may be supported
on a gantry, for example.
[0028] The apparatus 100 further comprises a controller 108 configured to control relative
movement between the tool 106 and the component 302. In this example, the controller
108 is configured to actuate the robotic arm 102 to move in order to effect relative
movement between the tool 106 supported by the tool mount 104 and a stationary work
platform 300 and component 302. However, in other examples, the controller 108 may
be coupled to a moveable work platform and/or a moveable apparatus 100 to control
relative movement therebetween.
[0029] The component 302 shown in Figure 1 is a block of metal, such as mild steel, stainless
steel, Inconel alloys and titanium, and the abrasive tool 106 comprises abrasive material
such as aluminium oxide. Figure 1 shows the apparatus 100 supporting the abrasive
tool 106 to engage an outer edge of the component 302, for example to machine a chamfer
onto the edge of the component 302.
[0030] The monitoring apparatus 200 comprises a controller 202 including a fuzzy inference
module 204 and a plurality of sensors. In this example, the sensors include an accelerometer
206, an acoustic emission sensor 208 and a force sensor 210. In the example arrangement
shown in Figure 1, the accelerometer 206 is coupled to the component 302 (for example,
it may be attached to the component 302 or the platform 300 by a magnetic mounting).
Similarly, the acoustic emission sensor 208 ("AE sensor") may be coupled to the component
by a magnetic mounting. A clamp, such as a G-clamp, may be used in place of a magnetic
mounting, particularly for non-magnetic components such as those comprising stainless
steel. The force sensor 210 is coupled to the tool mount 104.
[0031] In this example, the accelerometer 206 is a multi-axis accelerometer. In this particular
example, the accelerometer is a three-axis accelerometer, and the absolute magnitude
of the acceleration energy is determined from the accelerometer, rather than an acceleration
energy in a particular direction or plane. The acceleration energy is determined based
on a voltage output from the accelerometer (for example based on power spectral density
using a data acquisition module, or DAQ), and in this example is proportional to the
magnitude of acceleration.
[0032] The AE sensor 208 is configured to detect acoustic emission waves which are emitted
from within the component 302. Acoustic emission is a phenomenon by which acoustic
waves are generated within a solid article in response to forces or stresses occurring
within the article, for example during machining or other sources of stress. AE waves
therefore differ from conventional acoustic (i.e. sound waves) issued from sources
of sound. The AE sensor 208 may be configured to filter for AE waves based on filtering
a for a range of AE wave frequencies, such as between 1 kHz to 100MHz.
[0033] In this example, the force sensor 210 is a three-axis force sensor, and in this example
the force signal corresponds to (in particular, is proportional to) the magnitude
of the resultant force measured by the force sensor 210.
[0034] The controller 202 is configured to receive an acceleration signal from the accelerometer
206, an AE signal from the AE sensor 208 and a force signal from the force sensor
210. The fuzzy inference module 204 is configured to determine an abrasion parameter,
such as a chamfer dimension, surface roughness or material removal rate, based on
at least the acceleration signal and AE signal, as will be described in detail below
with respect to Figures 4 to 7.
[0035] Figure 2 shows a second example component 304 comprising a cuboidal block having
a cylindrical opening 306 extending therethrough. Figure 2 also shows an abrasive
tool 106 having a substantially cylindrical outer abrading surface 110 (defined by
the outer edges of a plurality of abrasive elements mounted on a hub of the tool 106)
and a shaft 112 which in use is supported and rotated by the tool mount. In this example,
the abrasive tool 106 is positioned to machine a chamfer onto one of the two circular
edges of the cylindrical opening 306.
[0036] The second example component 304 may be used for calibrating for abrasion monitoring,
as will be described in detail below with respect to Figure 8.
[0037] Figure 3 shows correlated acceleration and AE signals together with an output signal
for an abrasion parameter, in particular a chamfer dimension. The units of the acceleration
signal are m/s
2 (in examples, the acceleration energy may be expressed in units of g, i.e. 9.81m/s
2). The units of the AE signal in this example are voltage. In particular, a root mean
square, time-averaged AE signal is obtained based on the output of the voltage output
of the AE sensor. The chamfer dimension is shown in units of mm.
[0038] Each of the signals is shown against an X-axis corresponding to positions around
the inner circular edge of the component 304 of Figure 2. The units of the X-axis
correspond to six locations around the opening that are angularly separated by 60°.
In this particular example, a correlation can be observed between the acceleration
signal and the chamfer dimension when the chamfer dimension is relatively high (e.g.
at locations 4 to 6). Further, a correlation can be observed between the AE signal
and the chamfer dimension when the chamfer dimension is relatively low (e.g. at locations
1 to 3). The correlations may be difficult to observe and interpret based on each
individual signal, but the applicant has found that the fuzzy inference system (FIS)
process is capable of discerning the correlation and accurately predicting the output
variable (i.e. chamfer dimension), as will be described below.
[0039] The chamfer dimension is one example of an abrasion parameter that can be determined
by the controller 202. Other example abrasion properties include a rate of material
removal during abrasive machining, and a surface roughness, as will be described below.
[0040] A method of determining an abrasion parameter and of controlling an abrasive machining
operation will now be described with reference to Figures 4 to 7. For the purposes
of illustration only, the methods will be described with respect to the example abrasive
machining apparatus 100, abrasion monitoring apparatus 200 and component 304 described
above with respect to Figures 1 to 2.
[0041] Principles of a fuzzy inference method will first be described with respect to Figures
4 to 6, and a machining and monitoring method will be described with respect to Figure
7.
[0042] The principles of fuzzy inference systems have previously employed in optimisation
and control algorithms, for example in manufacturing.
[0043] Figure 4 shows a plot of example triangular-shape membership functions for an acceleration
signal. The X-axis represents the range of acceleration signal values that are expected
to be experienced in use, and in this example extends from 0g (i.e. 0m/s
2) to 0.15g (i.e. approximately 1.5m/s
2). The example triangular membership functions are designated VS (very small), S (small),
M (medium), L (large) and VL (very large) and extend over overlapping portions of
the range of acceleration values. Accordingly, a particular crisp value for the acceleration
signal (i.e. a single scalar value as sampled from an acceleration signal, such as
0.05g) may lie within two or more membership functions. For example, a crisp value
of 0.05g falls within both the S and M membership functions.
[0044] Truth values can be determined for each membership value. The truth value may correspond
to how closely the input variable (in this example a crisp value of the acceleration
signal) corresponds to a representative value in the membership function, such as
a central value. In the case of triangular-shape membership functions, a truth value
of between 0 and 1 may be determined based on relationships that define the shape
of the membership function. For example, a truth value of 1 is obtained for the S
membership function when the crisp value for acceleration is 0.045g, and the truth
value linearly decreases towards 0 for acceleration values increasing and decreasing
from 0.045g to limits at 0.06g and 0.03g respectively. Accordingly, it will be appreciated
that the truth value for an acceleration value of 0.05g will be higher for the S membership
function than for the M membership function.
[0045] Figure 5 shows a plot of example triangular-shape membership functions for the AE
signal which are similar to those for the acceleration signal.
[0046] The process of determining if an input variable falls within a membership function
may be referred to as fuzzification of the input variable.
[0047] Figure 6 shows a selection of a set of rules, for example 5 rules of a set of 25
rules, which can be evaluated based on the membership functions, to determine an output
variable Y, for example a chamfer dimension in units of mm.
[0048] For example, the first rule shown in Figure 6 specifies that if the AE signal is
VS (very small) and the acceleration signal is VS (very small), then the output variable
Y is VS (very small). As mentioned above, each of the signals may belong to more than
one membership function, and therefore several rules of the plurality of rules may
be satisfied.
[0049] The output variable Y may have similar membership functions, such as VS, S, M, L,
VL, to those described above with respect to the input variables of the AE signal
and the acceleration signal.
[0050] Accordingly, a plurality of rules within a set of rules may determine that the output
variable falls within the same or different membership functions. For example, two
rules may determine the output variable to be within the VS membership function, and
two further rules may determine the output variable to be within the S membership
function.
[0051] A de-fuzzification process is applied to determine a crisp (i.e. scalar) value for
the output variable, such as the chamfer dimension. In particular, a scalar value
for each rule may be determined based on the truth value or values of the input membership
functions, for example by mapping the lowest truth function of the input variables
(in the case of an AND logical definition) to the respective membership function and
determining a crisp value accordingly. The scalar values may then be averaged, for
example by a weighted average with different weights applying to each of the rules.
[0052] In other examples, the truth values may be used to perform a "centroid defuzzification"
based on overlapping membership functions for the output variable as is known in the
art, to determine a crisp value for the output variable.
[0053] The use of a fuzzy inference system as broadly described above enables quick and
computationally inexpensive evaluation based on rules which may be determined based
on empirical observations and optimisation, rather than by pure derivation from theoretical
engineering relationships which may be difficult to compute. Such evaluations may
be rapid to compute once the parameters of the respective rules are established.
[0054] In other examples, an output variable may be determined based on the input variables
without use of a fuzzy inference system. For example, a neural network may be provided
to map between the input variables and the output variable, or a database comprising
multi-dimensional look-up tables may be provided.
[0055] Figure 7 shows an example method 700 of abrasive machining. In block 702, instructions
for machining a component are defined. In this example, the instructions are defined
for abrasive machining of the component 304 to form a chamfer on the circular edge
of the opening 306 having a dimension of 1.5mm (i.e. the distance by which the chamfer
separates the cylindrical surface of the opening from the planar surface on the square
face of the component as shown in Figure 2). The instructions are defined so that
the abrasive tool 106, which in this example is a rotary compliant abrasive tool,
engages the edge to form a chamfer that is inclined by 45° with respect to each of
the two adjacent surfaces.
[0056] In this example, the instructions are defined so that the controller 108 causes the
rotary adjacent tool to traverse around the circular edge until the chamfer dimension
is 1.5mm around the entire edge. For example, the rotary abrasive tool may initially
traverse at a uniform speed of approximately 30mm/s, measured based on the path of
the rotating axis of the tool.
[0057] In block 704, the controller 108 causes a portion of the machining operation to be
conducted based on the machining instructions whilst the accelerometer 206 and AE
sensor 208 are generating the respective acceleration and AE signals respectively.
In this example, the portion represents a single sampling period for abrasion monitoring,
in particular 25ms. In other examples, a force sensor 210 and a respective force signal
may also be used.
[0058] The acceleration and AE signals (blocks 706, 708) are sampled by the controller 202
of the abrasion monitoring apparatus 200, for example at respective frequencies of
40kHz and 100kHZ, and based on the sampling over the 25ms sampling period an energy
calculation is performed for each signal to determine a crisp (i.e. scalar) acceleration
value relating to the energy of the acceleration signal over the sampling period,
and a crisp AE value relating to the energy of the AE signal over the sampling period.
[0059] In blocks 710 and 712, the fuzzy inference module 204 determines within which of
a plurality of respective membership functions defined for the fuzzy inference method
the respective crisp values lie. This step is referred to as fuzzification of the
input variables. For example, it may be determined that the crisp acceleration value
lies within both the VS and S membership functions with respective truth values of
0.7 and 0.25.
[0060] In block 714, a plurality of FIS (Fuzzy Inference System) rules as previously determined
are applied based on the fuzzy input variables by the fuzzy inference module 204.
In this particular example, there are 25 rules each operating on both an acceleration
fuzzy input variable and an AE fuzzy input variable. In other examples, some rules
may apply to only one input variable, or more than two input variables. In yet further
example methods, a force fuzzy input variable corresponding to the force acting between
the tool 106 and the component 304 may be provided, and different rules may be based
on combinations of one, two, three or more fuzzy input variables including fuzzy input
variables for acceleration, AE and force.
[0061] In block 716, the output variable of each of the FIS rules are combined, for example
based on a weighted average scheme using predetermined weights for each rule, and
an abrasion parameter is thereby determined (block 718) by the fuzzy inference module
204.
[0062] The abrasion parameter 718 is then used in a feedback loop for the abrasive machining
process as controlled by the controller 108. In this example, a feedback loop is implemented
to first determine whether machining is to be terminated (block 720), and if not,
determining whether the machining instructions are to be modified (block 722). The
feedback loop may be controlled by either the controller 108 of the abrasive machining
apparatus 100, by the controller 202 of the abrasion monitoring apparatus, a combination
of the two or by an integrated controller (i.e. controlling both machining and monitoring).
[0063] In example methods, the abrasion parameter (such as chamfer dimension) may be used
to determine an end point for an abrasive machining operation. For example, if machining
is being conducted at a single location, an abrasion parameter for that location can
be monitored and machining can be terminated (block 724) when the abrasion parameter
reaches a threshold value (such as a desired chamfer dimension or surface roughness).
[0064] When machining is being conducted over a plurality of locations, as in the present
example, it may only be appropriate to terminate the machining operation when machining
is already complete at all other locations.
[0065] In this particular example, machining is to be conducted around the perimeter of
the circular edge of the component 304 in a plurality of locations, and so instructions
to determine whether to terminate machining in block 720 are defined so that the machining
operation is only terminated when machining is determined to be completed in all locations,
as will be described in further detail below.
[0066] In this example, the abrasive tool 106 is controlled to traverse over the circular
edge at a uniform rate until machining is complete. The abrasion parameter is used
to monitor the progress of the machining operation by monitoring how the chamfer dimension
changes at a plurality of locations around the circular edge. For example, initially
the chamfer dimension may be highly variable around the circular edge, for example,
between 0mm and 0.5mm. Accordingly, different locations may require different amounts
of machining.
[0067] In block 722, the abrasion parameter is used to determine if the instructions for
the machining operation are to be modified. As described above, the instructions for
the machining operation are initially defined so that the abrasive tool moves around
the perimeter of the circular edge at a uniform rate. After at least one cycle of
the edge, a distribution of the abrasion parameter (in this example, the chamfer dimension)
is determined. In one example, the rate at which the abrasive tool is controlled to
traverse around the edge is adjusted based on the distribution of the abrasion parameter,
for example so that the abrasive tool 106 traverses at a slower rate over regions
with a relatively low chamfer dimension, and traverses at a faster rate over regions
with a relatively high chamfer dimension. In other examples, other control parameters
may be adjusted, such as a force at which the abrasive tool is applied against the
component (for example, the force may be selectively reduced or increased to adjust
the rate of abrasion).
[0068] In this particular example, the machining instructions are modified or redefined
once per revolution of the circular edge. Therefore, if a full revolution has not
yet been completed, then it is determined in block 722 that no modification of the
machining instructions is required and the method proceeds to conduct the next portion
of the machining operation in block 704. The loop continues accordingly. When a full
revolution is complete, it is determined at block 722 that modification of the machining
instructions is required, and the machining instructions are redefined in block 702
based on the feedback of the abrasion parameter around the perimeter of the circular
edge.
[0069] In other examples, the abrasive tool 106 may be configured to remain at a first location
until machining is complete at that location and before moving to a second location
(and so forth). Accordingly, in such examples the machining instructions may be redefined
in block 722 to cause movement to the second location only once machining is determined
to be completed at the first location, based on the abrasion parameter.
[0070] Whilst the above example has been described with respect to an abrasion parameter
which is a chamfer dimension, it will be appreciated that in other examples the abrasion
parameter may be different. For example, the abrasion parameter may be a surface roughness,
and the machining instructions may be defined so that abrasive machining continues
over a surface until all parts of the surface have a surface roughness below a predetermined
threshold roughness. Further, in other examples the abrasion parameter may be a rate
of material removal, and the machining instructions may be defined to achieve a predetermined
profile of material removal, for example based on a previous geometric analysis of
a component which identifies how much material is to be removed at different locations
on a component.
[0071] A method of calibrating a monitoring apparatus and/or a controller or machine-readable
instructions for such a monitoring apparatus will now be described with reference
to Figure 8.
[0072] For the purposes of illustration, the calibration method 800 will be described with
reference to the abrasion monitoring apparatus 200 of Figure 1 and the component 304
of Figure 2.
[0073] In block 802, machining instructions are defined to cause the abrasive tool 106 to
traverse one revolution around the circular edge of the opening 306 of the component
304. During machining, the acceleration signal 804 and the AE signal 806 are determined
and these are supplied as input variables to an FIS model 808 which is initially defined
based on a baseline set of rules. For example, the baseline set of rules may be defined
so that there is a rule for each permutation of the respective membership functions
of the input variables. For example, if there are five membership functions for each
input variable, and two input variables are used, then there will be 25 rules. The
baseline set of rules may initially be assumed to have an equal weighting.
[0074] In block 810, the abrasion parameter, which in this example is the chamfer dimension,
is predicted based on the FIS model 808 and the input variables.
[0075] In block 814, empirical abrasion parameter data is determined by inspection of the
component. For example, a laser measurement device may be used to determine the chamfer
dimension as is known in the art. In other examples the abrasion parameter may be
different and the empirical abrasion parameter data may be determined using alternative
tools accordingly. For example, when the abrasion parameter is surface roughness,
a profilometer may be used. Further, when the abrasion parameter is material removal
rate, an apparatus may be provided to capture material removed during machining to
determine the mass of material removed, for example by weighing the material.
[0076] In block 812, an error calculation is performed to determine the error between the
abrasion parameter predicted by the FIS model 808 and the empirical abrasion parameter.
The error may be provided to an optimisation sequence for the FIS rules, for example
based on genetic algorithms as will be briefly described below.
[0077] In block 818, the error is set as the objective function for a genetic algorithm
optimisation procedure for determining the weights of the rules. In block 820, the
fitness of a chromosome that determines the weights of the rules is determined based
on the objective function.
[0078] In block 822, it is determined if the fitness of the chromosome, based on the objective
function, is at a maximum. This may be determined to occur when there is no change
between a predetermined number of optimisation cycles. If the fitness is not at a
maximum, then in block 816 the chromosome is adapted based on genetic algorithm procedures
as are known in the art, to determine new weights for the FIS model 808.
[0079] A new FIS prediction 810 is thus determined and a new error calculation 812 is determined
based on a comparison with the pre-determined abrasion parameter data (i.e. without
repeating abrasive machining), which is again provided as the objective function for
optimisation.
[0080] By proceeding in this loop, the parameters for the FIS model (in particular the weighting
of the outcomes of the different rules) are determined by genetic algorithm optimisation,
to automatically arrive at a set of FIS parameters that calibrate the FIS model 808
based on the abrasion parameter data.
[0081] When it is determined that the fitness of the GA chromosome is at a maximum, the
chromosome is decoded (block 824) and the FIS model parameters (in particular the
weights for the respective rules) are stored.
[0082] Calibration may occur for each different abrasion parameter to be monitored, resulting
in different model parameters for each abrasion parameter. Further, calibration may
be conducted for different materials (component or abrasive), and different types
of tools.
[0083] However, the applicant has found that the FIS model performs consistently for different
geometry types. Accordingly, an FIS model calibrated for chamfer dimension based on
the component 304 can be applied for monitoring abrasion of other components having
different geometries.
[0084] By monitoring a plurality of parameters, in particular acceleration (of the component),
acoustic emission (AE) from within the component and optionally a force acting between
the component and an abrasive tool, the applicant has found that it is possible to
indirectly monitor an abrasion parameter such as surface roughness, a chamfer dimension
or a material removal rate. Accordingly, it becomes possible to monitor and adjust
an abrasive machining process, for example to modify an abrasion force, a path or
speed of an abrasion tool, or termination of abrasive machining based on feedback
from monitoring without interrupting machining, and/or without removing an abrasive
tool from a location on a component being monitored. Accordingly, machining times
may be reduced, accuracy may be improved, machining errors may be reduced (e.g. by
enabling continuous monitoring), and manual interaction with abrasive machining equipment
for periodic inspection may be reduced.
[0085] Further, such indirect monitoring may be of particular benefit for abrasive machining,
since abrasive tools may be compliant. Accordingly, it may not be possible to predict
abrasion parameters such as a chamfer dimension based on relative positioning of a
tool mount and a component alone, since the response of the compliant surface of the
abrasive tool, and therefore a modified profile of the component, may not be predictable.
[0086] Although none of the input variables described above (i.e. acceleration, acoustic
emission and force) have been shown to individually present a direct or linear correlation
with abrasion parameters such as surface roughness, chamfer dimension and material
removal rate, the applicant has found that their combination (in particular, of at
least acceleration and acoustic emission data), together with suitable optimisation
of FIS rules, results in reliable prediction independent of component geometry. By
way of example, Figure 9 shows FIS prediction vs empirical data for chamfer dimension
over a number of different passes of the component 304, which demonstrates an accurate
match. The applicant has found an accuracy of over 95% using this method.
[0087] Although examples have been described in which a method of abrasive machining is
controlled based on feedback from abrasion monitoring, it will be appreciated that
in examples, abrasion monitoring may be conducted in isolation to abrasive machining.
For example, abrasion monitoring may be conducted independently of the control of
an abrasive machining process. This may be useful, for example, if an abrasive machining
process is to be repeated for multiple different components, and may only practicably
be modified after machining of a particular component is complete. Therefore, abrasion
monitoring may provide useful data for modification of a repeatable process.
[0088] It will be appreciated that the methods described herein with respect to Figures
1 to 9 may be at least partly implemented in a computer, such as a general purpose
computer, or in an apparatus comprising a controller configured to implement the methods.
[0089] An example controller may include at least one processor and at least one memory.
The memory may store a computer program comprising computer readable instructions
that, when ready by the processor, causes the performance of at least one of the methods
described herein with respect to Figures 7 and 8. The computer program may be software
or firmware, or may be a combination of software and firmware.
[0090] The processor may include at least one microprocessor and may comprise a single core
processor, may comprise multiple processor cores (such as a dual core processor or
a quad core processor), or may comprise a plurality of processors (at least one of
which may comprise multiple processor cores).
[0091] The memory may be any suitable non-transitory computer (or machine) readable storage
medium, data storage device or devices, and may comprise a hard disk and/or solid
state memory (such as flash memory). The memory may be permanent non-removable memory,
or may be removable memory (such as a universal serial bus (USB) flash drive).
[0092] As shown in Figure 10, a non-transitory machine readable storage medium 1002 may
be provided including machine-readable instructions 1004 (or a computer program) executable
by a processor 1006 to cause performance of at least one of the methods described
herein with respect to Figures 7 and 8. The machine-readable instructions may be transferred
from the non-transitory machine-readable storage medium 1002 to a memory of a controller
or computer, such as the controller 202 of Figure 1. The non-transitory machine-readable
storage medium 1002 may be, for example, a USB flash drive, a compact disc (CD), a
digital versatile disc (DVD) or a Blu-ray disc. In some examples, the machine-readable
instructions may be transferred to a memory via a wireless signal or via a wired signal.
[0093] Further, the machine-readable instructions or computer program may be transmitted
by a signal that, when executed by a processor, causes the performance of at least
one of the methods described herein with respect to Figures 7 and 8.
[0094] It will be understood that the invention is not limited to the embodiments above-described
and various modifications and improvements can be made without departing from the
concepts described herein. Except where mutually exclusive, any of the features may
be employed separately or in combination with any other features and the disclosure
extends to and includes all combinations and sub-combinations of one or more features
described herein.
1. A method of monitoring abrasion of a component during abrasive machining, the method
comprising:
generating an acceleration signal corresponding to movement of the component or of
an abrasive tool in response to abrasive machining of the component by the tool;
generating an acoustic emission signal corresponding to acoustic waves emitted from
within the component;
determining an abrasion parameter relating to the removal of material from the component
based on a set of input variables including the acceleration signal and the acoustic
emission signal.
2. The method according to Claim 1, further comprising:
generating a force signal corresponding to a force acting between the abrasive tool
and the component; and
wherein the set of input variables includes the force signal.
3. The method according to Claim 1 or Claim 2, wherein the abrasion parameter is selected
from the group consisting of:
a material removal parameter relating to the rate at which material is removed from
the component by abrasion;
a chamfer dimension parameter relating to a dimension of a chamfer formed by abrasion;
and
a surface roughness parameter relating to a surface roughness of the component.
4. The method according to any preceding claim, wherein the abrasion parameter is determined
by a fuzzy inference module configured to determine the abrasion parameter by evaluating
a plurality of fuzzy logic rules based on the set of input variables.
5. A method of abrasive machining, comprising:
causing an abrasive tool to engage a component to conduct an abrasive machining operation;
monitoring abrasion of the component in accordance with any one of the preceding claims
to determine an abrasion parameter;
controlling the abrasive machining operation based on the abrasion parameter.
6. The method according to Claim 5, wherein at least one of a tool engagement force;
a path of relative movement of the abrasive tool over the component; and a residence
time of the tool at a first location of the component is controlled based on the abrasion
parameter.
7. The method according to Claim 5 or Claim 6, wherein controlling the abrasive machining
operation comprises comparing a value of the abrasion parameter for a first location
of the component determined by monitoring with a target value for the first location.
8. The method according to any one of Claims 5 to 7, wherein the abrasive machining operation
is selected from the group consisting of:
linishing;
deburring;
surface grinding; and
edge grinding.
9. An apparatus comprising:
a tool mount for supporting an abrasive tool to engage a component;
an accelerometer for generating an acceleration signal corresponding to movement of
the component or movement of the abrasive tool in response to abrasive machining of
the component by the tool;
an acoustic emission sensor for generating an acoustic emission signal corresponding
to acoustic waves emitted from within the component in response to abrasive machining
of the component by the tool; and
a controller configured to determine an abrasion parameter relating to the removal
of material from the component based on a set of inputs including the acceleration
signal and the acoustic emission signal.
10. The apparatus according to Claim 9, wherein the accelerometer and/or the acoustic
emission sensor is configured to be coupled to the component.
11. The apparatus according to Claim 9 or Claim 10, further comprising:
a force sensor coupled to the tool mount and configured to output a force signal corresponding
to a force acting between an abrasive tool supported by the tool mount and a component
with which it is engaged in use;
wherein the controller is configured to determine the abrasion parameter based on
a set of inputs including the force signal.
12. The apparatus according to any of Claims 9 to 11, wherein the tool mount is for supporting
a rotary abrasive tool.
13. The apparatus according to any one of Claims 9 to 12, further comprising a compliant
abrasive tool supported by the tool mount.
14. A non-transitory machine-readable storage medium encoded with instructions executable
by a processor, including instructions to:
receive an acceleration signal corresponding to movement of a component or movement
of an abrasive tool in response to abrasive machining of the component by the tool;
receive an acoustic emission signal corresponding to acoustic waves emitted from within
the component during abrasive machining of the component;
determine an abrasion parameter relating to the removal of material from the component
based on a set of inputs including the acceleration signal and the acoustic emission
signal.
15. The non-transitory machine-readable storage medium according to Claim 14, including
instructions to:
receive a force signal corresponding to a force acting between the abrasive tool and
the component during abrasive machining; and
wherein the set of inputs includes the force signal.